This morning I was watching an FT interview with Demis Hassabis of Google DeepMind, and it made me reflect on the different approaches to AI between China and the US.

In China, the race is not primarily about building the best LLM model. The focus is much more on applications.

In other words, the competition happens on the last layer of the cake, not on the ingredients.

The North Star is not necessarily building AGI, but deploying AI power in everyday use cases as quickly and broadly as possible.

By contrast, the US — and often Europe — tends to follow a more centralized model, focusing heavily on the core technology itself: the models. The emphasis is on improving the ingredients rather than experimenting with the final recipe.

This creates a paradox: US seems to centralize, wherever China prefers open systems?

China’s ecosystem appears far more application-driven, with an enormous appetite for open access models and tools — platforms that allow developers and companies to build their own engines and solutions directly.

Meanwhile, many Western AI leaders seem to prefer a different path. From the interview, it feels as if companies like DeepMind are leaning toward partnering with industries and providing controlled access to their models, rather than fully opening the technology and letting everyone build on top of it independently.

Two very different philosophies are emerging:

The interesting question is which strategy will ultimately create the most economic value.

- SL